News of L’Oréal’s partnership with OpenAI sent shock waves through the beauty industry this summer, turning the spotlight on the need for new ways to tackle the consumer journey in the age of AI.
The beauty powerhouse has teamed up with the tech giant to provide enriched content to power AI searches. There is a Maybelline x Modiface try-on app inside ChatGPT, for example, and a U.S. pilot of a verified product feed for which L’Oréal supplies Lancôme and Kérastase stock-keeping-unit descriptions to ChatGPT.
The deal comes as AI is fundamentally altering how consumers search for product recommendations. People typically spend three to four minutes browsing products in a traditional Google search. But with AI’s more iterative approach, a “conversation” stretches to roughly 11 minutes per Google AI session, according to Google Analytics and eMarketer AI Mode Session Tracking. (L’Oréal calls this “the 11-minute paradox.”)
The information comes at a time of mass migration to AI usage. OpenAI’s ChatGPT has more than 1 billion monthly active users, or MAU. Google AI Overviews counts over 2.5 billion MAU, while the purely LLM app Google Gemini has 900 million-plus MAU.
That migration to AI is evident in the beauty space. Fifty-five percent of beauty consumers say they are weekly users of generative AI. That is 1.5 times more than in 2025, according to Accenture’s recent Consumer Pulse Research. For Gen Zers, the level rises to 76 percent. And over the past 12 months, AI became the top-ranked beauty discovery channel for weekly AI users, Accenture found.
That means that LLMs are now the front door to beauty discovery. Therefore with GEO, or generative engine optimization, content needs to be adjusted so a brand or product appears in the answer layer during a consumer’s search.
L’Oréal studied how shoppers discover, have deep consumer experiences after discovery, and how they buy. As people chat with AI online, how do L’Oréal brands appear in the answer layer?
Product content and data are key, as are descriptors. They all need to be enhanced for LLM versus traditional search. The overall aim is to make a customer’s conversation with AI better, more human and with the right information for a specific conversation.
“The way you produce content [for search] today is evolving versus how you produced content even two, three years ago,” said Audrey Depraeter-Montacel, Accenture’s global beauty industry lead. “Creativity is still absolutely needed, even more than ever.”
Having L’Oréal — the world’s largest beauty company — influencing the overall AI algorithm could put smaller brands at a disadvantage. But indies’ agility and their being less risk-averse than strategics are positive points of difference, some industry experts believe. Especially for indie brands, the challenge is how they appear in AI, which more frequently highlights legacy brands today. Then, once indies appear, what does that look like?
“With AI, because it works on frequency so much — meaning how often does this brand show up for this particular topic — you already have to be in all of those different [digital platforms] before you get a good probability of showing up,” said David Van Veen, managing director of beauty branding agency MSLK. “There starts to be a big alignment between PR functionality and search-engine optimization functionality.”
The right positioning becomes paramount. Messaging must be consistent and problem-solving-oriented. Follow-up prompts — like “this is the best foundation for me because of the amount of coverage I want for my skin type” — is where the real personalization comes in. The more a product can be connected to identity or a use case, the better.
“The content that exists right now for most brands, including indies, is not based around going really deep into those specifics,” Van Veen said.
Many of the most popular beauty brands on LLMs are clinically driven. “So brands really will need to show their claims, showcase what the product performance is,” said Anastasia Georgievskaya, chief executive officer and cofounder of Haut.AI.
In early August, Haut.AI launched Clinical Studies Software, billed to let beauty, ingredient and dermatology companies run remote and hybrid AI-enabled clinical research with hundreds to thousands of people and generate evidence at scale. The software measures 48 biomarkers and reduces the setup time of a study by more than 90 percent, according to Haut.AI.
The software generates a full report about product performance. It can describe with great granularity what the product improves. “Once it becomes more available to generate claims about products, this can almost become a strategy for brands to explain better how this product works, which is benefiting the customer and also benefiting their LLM search,” Georgievskaya said. “The more clinical studies you have, the more clinical data you have, it’s quite likely that LLM will pick up this information about your published study and show you better inside an LLM.”
While this kind of clinical study will often become public domain, companies that want to keep their product information and customer information private are building private data sets for company-specific apps and agents.
Haut.AI technology, such as its new Body Analysis, helps brands collect first-party data about their customers, which can help them align their messaging and types of content they put online to help consumers better find their products. Body Analysis not only analyzes skin health but also allows people to receive personalized recommendations.
Wayne Liu, president and chief growth officer of Perfect Corp., believes AI will revolutionize beauty with this kind of personalized, brand-specific digital intelligence. “At some point in time, you do need to create a personalized agent for your brand,” he said.
Liu is a proponent of using vertical data — including a brand’s social media feed, all its posts and customer survey data — to train a brand’s beauty agent, which in turn will help people discover products best suited to them.
He likened the AI landscape of today to the Wild West of yesteryear. “If you want to create your own ranches, you need to build a fence,” Liu said.